A picture is worth a thousand words : content-based image retrieval techniques

In my dissertation I investigate techniques for improving the state of the art in content-based image retrieval. To place my work into context, I highlight the current trends and challenges in my field by analyzing over 200 recent articles. Next, I propose a novel paradigm called ‘artificial imagination’, which gives the retrieval system the power to imagine and think along with the user in terms of what she is looking for. I then introduce a new user interface for visualizing and exploring image collections, empowering the user to navigate large collections based on her own needs and preferences, while simultaneously providing her with an accurate sense of what the database has to offer. In the later chapters I present work dealing with millions of images and focus in particular on high-performance techniques that minimize memory and computational use for both near-duplicate image detection and web search. Finally, I show early work on a scene completion-based image retrieval engine, which synthesizes realistic imagery that matches what the user has in mind.